1. SSAS - Designing, Development and Deployment Best practices Amit Bansal www.PeoplewareIndia.com www.SQLServerGeeks.com

2. About Amit Bansal CTO, eDominer Systems & Peopleware India Corporate Trainer/Consultant & Evangelist Conducted more than 400 workshops on SQL Server & BI for top notch IT companies world wide Microsoft MVP for SQL Server Microsoft Certified Trainer Advisory Council member Speaker at TechED India, TechED US & TechED Europe Technical Reviewer – MSL courses on SQL Server SME – SQL Server 2008 certifications President – SQLServerGeeks.com

3. Agenda Problem Statement Dimension Design Cube Design Partitioning Aggregations Summary

4. Problem Statement You want: –Faster initial development –Easier further development –Easier maintenance –Agility and scalability in your design –Performance, Performance, Performance You need to design best, right from start ! (Do you I really need to tell you this )

5. Agenda Problem Statement Dimension Design Cube Design Partitioning Aggregations Summary

6. Natural Hierarchies A hierarchy is a natural hierarchy when each attribute included in the user-defined hierarchy has a one to many relationship with the attribute immediately below it (every child member has only one parent) Server simply “works better” City State Country Month Quarter Year Date

7. Natural Hierarchies Performance implications –Only natural hierarchies are materialized on disk during processing –Unnatural hierarchies are built on the fly during queries (and cached in memory) –Server internally decomposes unnatural hierarchies into natural components –Essentially operates like ad hoc navigation path (but somewhat better) –Aggregation designer favors user defined hierarchies

8. Attribute Relationships Customer City State Country GenderMarital Age Illustration

9. Attribute Relationships Flexible relationships can change Rigid relationships do not change City State Phone Number Customer City Month Quarter

10. Attribute Relationships Where are they used? –Storage Query performance –Greatly improved effectiveness of in-memory caching –Materialized hierarchies when present Processing performance: Fewer, smaller hash tables result in faster, less memory intensive processing Aggregation design: Algorithm needs relationships in order to design effective aggregations Member properties: Attribute relationships identify member properties on levels

11. Attribute Relationships Where are they used? –Semantics MDX overwrite semantics: City.Seattle  State. WA | State.OR  City.All Non-key granularity (Aggregation Paths) Dimension security: DeniedSet = {State.WA}

12. DEMO Attribute relationships & Natural hierarchies

13. Dealing with Large Dimensions Optimizing Processing –Use natural hierarchies Good attribute/hierarchy relationships forces the AS engine to build smaller DISTINCT queries versus one large and expensive query Consider size of other properties/attributes –Dimension SQL queries are in the form of select distinct Key1, Key2, Name, …, RelKey1, RelKey2, … from [DimensionTable]

14. Dealing with Large Dimensions Important to tune your SQL statements –Indexes to underlying tables –Create a separate table for dimensions –Avoid OPENROWSET queries –Use Views to create your own version of “query binding” Size limitations for string stores and effect on dimension size –4 GB, stored in Unicode, 6 byte per-string overhead. –E.g. 50-character name: 4*1024*1024*1024 / (6+50*2) = 40.5 million members

15. Dimension Processing ByAttribute vs ByTable –This is a ProcessingGroup property Default = ByAttribute –Advantages of ByTable Entire set of dimension data loaded into memory Theoretically processes data faster But BEWARE –Bypasses normal checks –Assumes there is enough memory to process all attributes concurrently –If this is not true…

16. Dimension Processing ByAttribute vs ByTable –2 dimensions Each >25M members with 8-10 attributes – ByTable Took 80% of available memory 25.6 / 32 GB Never completed – ByAttribute Only 28% of available memory 9 / 32 GB Process completed

17. Agenda Problem Statement Dimension Design Cube Design Partitioning Aggregations Summary

18. Cube Dimensions Dimensions –Consolidate multiple hierarchies into single dimension (unless they are related via fact table) –Use role-playing dimensions (e.g., OrderDate, BillDate, ShipDate)—avoids multiple physical copies –Use parent-child dimensions prudently No aggregation support –Set Materialized = true on reference dimensions

19. Cube Dimensions Dimensions –Use many-to-many dimensions prudently Slower than regular dimensions, but faster than calculations Intermediate measure group must be “small” relative to primary measure group Consider creating aggregations on the shared common attributes of the intermediate measure group

20. Measure Groups Common questions –At what point do you split from a single cube and create one or more additional cubes? –How many is too many? Why is this important? –New measure groups adding new dimensions result in an expansion of the cube space –Larger calculation space = more work for the engine when evaluating calculations

21. Measure Groups Guidance –Look at increase in dimensionality. If significant, and overlap with other measure groups is minimal, consider a separate cube –Will users want to analyze measures together? –Will calculations need to reference unified measures collection?

22. DEMO Cube Design Best Practices

23. Agenda Problem Statement Dimension Design Cube Design Partitioning Aggregations Summary

24. Why Partition? Breaks large cubes into manageable chunks For measure groups, not dimensions Fact rows are distributed by a partitioning scheme –Managed by DBA –By Time: Sales for 2001, 2002, 2003, … –By Geography: Sales for North America, Europe, Asia, … Why? –For Manageability, Performance, Scalability

25. Benefits of Partitioning Partitions can be added, processed, deleted independently –Update to last month’s data does not affect prior months’ partitions –Sliding window scenario easy to implement –e.g., 24 month window  add June 2006 partition and delete June 2004 Partitions can have different storage settings –Storage mode (MOLAP, ROLAP, HOLAP) –Aggregation design –Alternate disk drive –Remote server

26. Benefits of Partitioning Partitions can be processed and queried in parallel –Better utilization of server resources –Reduced data warehouse load times Queries are isolated to relevant partitions  less data to scan –SELECT … FROM… WHERE [Time].[Year].[2006] –Queries only 2006 partitions Bottom line  partitions enable –Manageability, Performance & Scalability

27. Best Practices for Partitions General guidance: 20M rows per partition –Use judgment, e.g., perhaps better to have 500 partitions with 40 million rows than 1000 20 million row partitions –Standard tools unable to manage thousands of partitions More partitions means more files –E.g. one 10GB cube with ~250,000 files (design issues) –Deletion of database took ~25min to complete Partition by time plus another dimension e.g. Geography –Limits amount of reprocessing –Use query patterns to pick another partitioning attribute When data changes –All data cache for the measure group is discarded –Separate cube or measure groups by “static” and “real-time” analysis

28. Best Practices for Partitions January 2008 1120112021302130314031404150415011151115162016202125212526502650 Equal Sized Partitions Not Equal Sized Partitions

29. DEMO Partitioning

30. Agenda Problem Statement Dimension Design Cube Design Partitioning Aggregations Summary

31. Aggregations for query performance A subtotal of partition data –based on a set of attributes from each dimension Customers All Customers Country State City Name Products All Products Category Brand Item SKU Facts custIDSKUUnits SoldSales 345-231351232$45.67 563-0145123634$67.32 … Highest-Level Aggregation CustomerProductUnits SoldSales All 347814123$345,212,301.30 Intermediate Aggregation countryCodeproductIDUnits SoldSales Cansd4529456$23,914.30 USyu6784623$57,931.45 …

32. How many Aggregations  125 possible combinations (just for user-defined dimensions)  5 customer levels, 5 product levels, 5 time levels  Imagine a cube with ten dimensions, five levels each  9,765,625 combinations! Then you add attribute hierarchies to the mix  General rule: multiply the number of attributes in each dimension  Goal should be to find the best subset of this potentially huge number of possibilities Tradeoff between query performance and processing/storage overhead Customer All Customers CountryStateCityNameProduct All Products CategoryBrandItemSKU Time All Time YearQuarterMonthDay

33. Aggregations for query performance Customers All Customers (1) Country (3) State (80) City (578) Name (3811) Product All Products (1) Category (60) Brand (911) Item (7621) SKU (8211) Time All Time (1) Year (3) Quarter (12) Month (36) Day (1095) Aggregations at lower levels have more possible rows… (All, All, All)1 x 1 x 1= 1 (Country, Item, Quarter) 3 x 7621 x 12= 274,356 (Name, SKU, Day)3811 x 8211 x 1095= 34,264,872,495 Actual number of rows depends on the data sparsity Size also depends on the number of measures

34. Aggregations for query performance Query levelsAggregation used Max Cells (All, All, All)(All, All, All) 1 (Country, Item, Quarter)(Country, Item, Quarter) 274,356 (Country, Brand, Quarter)(Country, Item, Quarter) 274,356 (Country, Category, All) (Country, Item, Quarter) 274,356 (State, Item, Quarter)(Name, SKU, Day) 34,264,872,495 (City, Category, Year) (Name, SKU, Day) 34,264,872,495 Using a higher-level aggregation means fewer cells to consider Customers All Customers Country State City Name Products All Products Category Brand Item SKU Time All Time Year Quarter Month Day

35. Best Practices for Aggregations Define all possible attribute relationships Set accurate attribute member counts and fact table counts Set AggregationUsage –Set rarely queried attributes to None –Commonly queried attributes to Unrestricted

36. Best Practices for Aggregations Not too many –In the 100s, not 1000s! Do not build aggregations > 30% of fact table size

37. Best Practices for Aggregations 1.Use Storage Design Wizard for the initial aggregations (~20% perf gain) 2.Enable query log 3.Run pilot workload with limited users 4.Refine with Usage Based Optimization Wizard 5.Use a larger perf gain (70+%) 6.Reprocess partitions for new aggregations to take effect 7.Periodically use UBO to refine aggregations

38. DEMO Aggregations

39. Agenda Problem Statement Dimension Design Cube Design Partitioning Aggregations Summary

40. Design for performance and scalability from the start Some fundamental principles carry through from SQL 7.0 –Dimension design –Partitioning –Aggregations Critical to properly implement/utilize modeling capabilities introduced in SSAS 2005 and carried forward in 2008 –Attribute relationships, natural hierarchies –Design alternatives: role-playing, many-to-many, reference dimensions, semi-additive measures –Flexible processing options SSAS 2008 development tools have been redesigned and enhanced to better assist in development of high performance cubes

41. Resources Analysis Services 2005 Processing Architecture http://msdn.microsoft.com/en-US/library/ms345142(v=SQL.90).aspx Many-to-Many Dimensions in Analysis Services http://msdn.microsoft.com/en-US/library/ms345139(v=SQL.90).aspx Analysis Services Query Performance Top 10 Best Practices http://msdn.microsoft.com/en-US/library/cc966527.aspx SQL Server 2008 Analysis Services Performance Guide http://msdn.microsoft.com/en-us/library/dd542635(v=SQL.100).aspx

42. Resources Software Application Developers http://msdn.microsoft.com/ Infrastructure Professionals http://technet.microsoft.com/ msdnindia technetindia @msdnindia @technetindia

43. © 2011 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries. The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.